demo.git - Vide-coded 64k demo system

Age	Commit message (Expand)	Author
13 hours	Fix CNN v2 static feature channel mapping (p4-p7 → channels 8-11)	skal
13 hours	gen_identity_weights: Change --mix to 50-50 blend	skal
13 hours	gen_identity_weights: Add --p47 option for static feature visualization	skal
13 hours	gen_identity_weights: Add --mix option for static feature blending	skal
14 hours	CNN v2: Fix Layer 0 visualization scale (was 0.5, now 1.0)	skal
14 hours	CNN v2: Add debugging tools for mismatch investigation	skal
14 hours	CNN v2 training: Fix float64/float32 dtype mismatch in depth feature	skal
14 hours	CNN v2: Alpha channel depth handling and layer visualization	skal
15 hours	CNN v2: Use alpha channel for p3 depth feature + layer visualization	skal
20 hours	CNN v2 training: Add --grayscale-loss option for luminance-based loss computa...	skal
21 hours	CNN v2: Change feature #6 from sin(10x) to sin(20y)	skal
21 hours	CNN v2: Add TODO for flexible feature layout in binary format v3	skal
21 hours	CNN v2: Add mip-level support to runtime effect	skal
21 hours	CNN v2 export: Read and display mip_level from checkpoints	skal
21 hours	CNN v2: Add --mip-level option for parametric features	skal
22 hours	CNN v2: Fix activation function mismatch between training and inference	skal
25 hours	CNN v2 training: Use target image alpha channel	skal
25 hours	CNN v2: Restore per-layer kernel sizes support	skal
25 hours	CNN v2: Refactor to uniform 12D→4D architecture	skal
29 hours	Add weights/ subdirectory to workspaces for CNN training outputs	skal
47 hours	test_demo: Add beat-synchronized CNN post-processing with version selection	skal
2 days	Refine training script output and validation	skal
2 days	TODO: 8-bit weight quantization for 2× size reduction	skal
2 days	CNN v2: Storage buffer complete - real weights exported	skal
2 days	CNN v2: storage buffer architecture foundation	skal
2 days	TODO: Add random sampling to patch-based training	skal
2 days	CNN v2: Patch-based training as default (like CNN v1)	skal
2 days	Fix: CNN v2 training - handle variable image sizes	skal
2 days	CNN v2: parametric static features - Phases 1-4	skal
2 days	remove more stale files	skal
3 days	feat: implement beat-based timing system	skal
3 days	add trained layers	skal
3 days	docs: Update CNN comments and add bias fix summary	skal
3 days	fix: CNN bias accumulation and output format improvements	skal
3 days	update cnn code	skal
3 days	refactor: Use linspace(-1,1) directly for coords	skal
3 days	fix: Compute gray from [0,1] RGB in CNN shader generator	skal
3 days	fix: Complete auxiliary texture initialization fix	skal
3 days	add --save-intermediates to train.py and cnn_test	skal
3 days	fix: Move sigmoid activation to call site in CNN layer shader	skal
3 days	fix: Replace clamp with sigmoid in CNN final layer	skal
3 days	feat: Add early stopping to CNN training	skal
3 days	fix: CNN training/inference to match WGSL sliding window	skal
3 days	format .wgsl layer code (cosmetics)	skal
4 days	fix: Use patch-based inference to match CNN training distribution	skal
4 days	opt: Move invariant in1 calculation outside CNN convolution loops	skal
4 days	opt: Vec4-optimize CNN convolution shaders for SIMD	skal
4 days	chore: Update CNN architecture to 3×3×3 with new trained weights	skal
4 days	docs: Update CNN training documentation with patch extraction	skal
4 days	feat: Add salient-point patch extraction for CNN training	skal